This model performs English air-traffic-control automatic speech recognition.
It was trained using CTC and supervised contrastive learning with a hybrid
training schedule that combines:
The included processor supports greedy CTC decoding.
This model is intended for research on robust English ATC speech recognition.
It may be used to transcribe English ATC recordings with acoustic and speaker
characteristics similar to its training data.
The model was trained using UWB-ATCC speech and synthetic speech derived from
UWB-ATCC transcripts.
Users should review and comply with the original dataset terms before using or
redistributing this model.
1import torch
2import soundfile as sf
3from transformers import AutoModelForCTC, AutoProcessor
4
5model_id = "thaivanphat95/wav2vec2-robust-uwb-supcon-hybrid"
6
7processor = AutoProcessor.from_pretrained(model_id)
8model = AutoModelForCTC.from_pretrained(model_id).eval()
9
10audio, sample_rate = sf.read("audio.wav")
11inputs = processor(audio, sampling_rate=sample_rate, return_tensors="pt")
12
13with torch.no_grad():
14 logits = model(**inputs).logits
15
16prediction_ids = torch.argmax(logits, dim=-1)
17transcript = processor.batch_decode(prediction_ids)[0]
18print(transcript)
Audio should be mono. Resample audio to 16 kHz before inference when necessary.
1@article{thai2026contrastive,
2 title={Contrastive Regularization for Accent-Robust ASR},
3 author={Thai, Van-Phat and Dhruv, Aradhya and Pham, Duc-Thinh and Alam, Sameer},
4 journal={arXiv preprint arXiv:2605.03297},
5 year={2026},
6 doi={10.48550/arXiv.2605.03297}
7}
The model weights are not covered by the Apache License 2.0 used for the
training code. Their use and redistribution may also be affected by the
licenses and terms of the pretrained model, UWB-ATCC, source transcripts, and
synthetic speech-generation systems.
